def test_anneal_minimize_raises_when_wrong_cooling_factor():
with pytest.raises(ValueError):
optimization.anneal_minimize(
'initial',
lambda s: 1.0 if s == 'initial' else 0.0,
lambda s: 'better',
lambda: 1.0,
1.0,
0.5,
2.0,
1,
)
def search(
self,
trace_func: Callable[[List[LineSequence], float, float, float, bool],
None] = None
) -> List[LineSequence]:
"""Issues new linear sequence search.
Each call to this method starts new search.
Args:
trace_func: Optional callable which will be called for each simulated
annealing step with arguments: solution candidate (list of linear
sequences on the chip), current temperature (float), candidate cost
(float), probability of accepting candidate (float), and acceptance
decision (boolean).
Returns:
List of linear sequences on the chip found by this method.
"""
def search_trace(state: _STATE, temp: float, cost: float,
probability: float, accepted: bool):
if trace_func:
trace_seqs, _ = state
trace_func(trace_seqs, temp, cost, probability, accepted)
seqs, _ = optimization.anneal_minimize(self._create_initial_solution(),
self._quadratic_sum_cost,
self._force_edges_active_move,
self._rand.random_sample,
trace_func=search_trace)
return seqs
def test_anneal_minimize_calls_trace_func():
trace_func = mock.Mock()
optimization.anneal_minimize('initial',
lambda s: 1.0 if s == 'initial' else 0.0,
lambda s: 'better',
lambda: 1.0,
1.0,
0.5,
0.5,
1,
trace_func=trace_func)
trace_func.assert_has_calls([
mock.call('initial', 1.0, 1.0, 1.0, True),
mock.call('better', 1.0, 0.0, 1.0, True)
])
def test_anneal_minimize_keeps_when_worse_and_discarded():
assert (
optimization.anneal_minimize(
'initial',
lambda s: 0.0 if s == 'initial' else 1.0,
lambda s: 'better',
lambda: 0.9,
1.0,
0.5,
0.5,
1,
)
== 'initial'
)
def test_anneal_minimize_improves_when_better():
assert (
optimization.anneal_minimize(
'initial',
lambda s: 1.0 if s == 'initial' else 0.0,
lambda s: 'better',
lambda: 1.0,
1.0,
0.5,
0.5,
1,
)
== 'better'
)